Namespaces and EKS

Hi all

So I see EKS creates a elastic-system namespace during the CRD creations.
So question, would/should we deploy our own cluster bits into elastic-system ns...
when you follow the quickstart they don't mention namespaces, implying your cluster ends in the default ns.
so option #2, better is to rather push it into the elastic-system ns... but I'd rather think option #3 is better, push the cluster into say elastic-prod ns... thus allowing you for whatever reason, (maybe testing, or multi tenancy) to have a second cluster in say elastic-test ns.
and thus leave the operator alone/in elastic-system,

Comment ?


The intention behind the elastic-system namespace is indeed to be reserved exclusively for the operator to isolate it from accidental deletion etc. Similar in spirit to how the Kubernetes control plane services are deployed in the kube-system namespace.

For your workloads (Elasticsearch clusters, Kibanas etc) you can chose whatever namespace you see fit which is why we do not include namespaces in our examples.

Hi Perbrc

Thanks for the confirmation, I would propose though, like allot of other technology/examples that do mention chose a namespace of your choice, to mention it, otherwise noobs will end deploying into default, until they've figured out it's a bad idea and then end redoing it, which is just not needed to start with.



I raised an issue to highlight our intention behind the elastic-system namespace in the docs.

Deploying into the default namespace is fine if you are just starting to use ECK and are still in the process of evaluation and experimentation. For anything production grade I would assume that working with namespaces in Kubernetes will no longer be a issue, but we can look into adding a sentence to the docs.

1 Like

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.